Why Indian IT is Not Keen on Building AI Foundational Models

While Indian startups and founders are catching up to the AI wave, prompted by the announcement of the $500 billion Stargate Project from the US and China’s open-source “side project” DeepSeek, the Indian IT sector remains unfazed. With unwavering confidence in Nandan Nilekani’s vision, they continue to focus on building India as the AI use case capital of the world.

Though tech giants like TCS, Infosys, Wipro, and HCLTech have started developing agentic AI frameworks, small language models, and even drug discovery, their efforts remain focused on clients alone. These initiatives do not prioritise building foundational technologies for the country. 

For example, Tech Mahindra built its Project Indus, the only foundational model emerging from an IT firm in India. However, the expected impact is largely on its clients instead of the creation of transformative products like ChatGPT, Claude, or Gemini. The model, uploaded on Hugging Face by Tech Mahindra chief innovation officer Nikhil Malhotra, is open-source. However, it has garnered very few downloads and is not available for Serverless Inference with its API.

Unfortunately, Indian IT is not very interested in building a foundational model. One can argue that even though the firms have enough funds to make a foundational model, they won’t build one unless their clients ask them to or there is some requirement from their side.

What Needs to be Done?

When AIM asked industry experts if India should also build its own Stargate Project, the responses were mostly positive. Ajai Chowdhry, HCL co-founder and chairman of the Mission Governing Board of India’s National Quantum Mission, expressed concern over the growing shift towards taking control of AI. 

“We seem to be getting to the weaponisation of tech. For strategic autonomy, we must create our own AI doctrine and have strong control over our data,” Chowdhry told AIM. 

Several others expressed similar thoughts as soon as Aravind Srinivas, CEO of Perplexity, said that India should definitely focus on building its own LLM and that Nilekani was “wrong about India not requiring it to build one.”

The co-founders of Infosys are still debating whether they need one. One of them, Kris Gopalakrishnan, wrote on X that India needs to build its foundation model for a cultural and strategic economy. 

Meanwhile, Mohandas Pai, the former CFO of Infosys, earlier told AIM that Indian IT companies are services companies and are not focused on building AI products. According to him, the funding required to build something foundational is much higher than what is available in the country.

Pai called for a government-backed innovation fund similar to France’s $36 billion France Innovation Fund, which supports startups like Mistral. Such funding, he argued, could enable India to produce foundational models and compete on a global scale.

This is similar to what K Krithivasan, CEO and MD of TCS, highlighted when he said that building LLMs has no huge advantage, as the cost outweighs the benefit. He added that since most organisations in India are system integrators, companies need to use products as software and ensure that clients receive the benefits.

At the same time, he also agreed that building it for regional languages makes sense for the democratisation of the technology.

CP Gurnani, former CEO of Tech Mahindra, told AIM that building a foundational model is important, and that is what led him to start the Project Indus initiative during his tenure. He said that India should build something like NVIDIA. 

How is the Work Progressing?

While AI startups like Sarvam and Krutrim are working with IT giants to provide language models for their clients, the firms are also interested in helping their clients build the models. However, instead of building one, it seems like the ideal approach for IT firms would be to provide a DeepSeek R1 model for their clients.

“The Indian path in AI is different. We are not in the arms race to build the next LLM; let people with capital, let people who want to pedal chips do all that stuff…We are here to make a difference, and our aim is to put this technology in the hands of people,” Nilekani said last year, back when the debate around building AI had started. 

The upskilling of the Indian IT workforce for generative AI is also a good sign for how the firms are adopting generative AI, but nothing about building a foundational model. It seems like Indian IT has missed the generative AI bus, and no amount of funds can bring it back up. 

As Narayana Murthy said last year, India is only good at copying ideas from the West and applying them here. India’s tech sector continues to prioritise short-term gains from outsourced IT services rather than investing in creating globally competitive products. For Indian IT, it actually makes sense. 

The post Why Indian IT is Not Keen on Building AI Foundational Models appeared first on AIM Media House.